2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "config-host.h"
25 #include "qemu-common.h"
27 #include "block_int.h"
29 #include "qemu-objects.h"
32 #include <sys/types.h>
34 #include <sys/ioctl.h>
35 #include <sys/queue.h>
45 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
46 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
47 BlockDriverCompletionFunc
*cb
, void *opaque
);
48 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
49 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
50 BlockDriverCompletionFunc
*cb
, void *opaque
);
51 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
52 BlockDriverCompletionFunc
*cb
, void *opaque
);
53 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
54 uint8_t *buf
, int nb_sectors
);
55 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
56 const uint8_t *buf
, int nb_sectors
);
57 static BlockDriver
*find_protocol(const char *filename
);
59 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
60 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
62 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
63 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
65 /* If non-zero, use only whitelisted block drivers */
66 static int use_bdrv_whitelist
;
68 int path_is_absolute(const char *path
)
72 /* specific case for names like: "\\.\d:" */
73 if (*path
== '/' || *path
== '\\')
76 p
= strchr(path
, ':');
82 return (*p
== '/' || *p
== '\\');
88 /* if filename is absolute, just copy it to dest. Otherwise, build a
89 path to it by considering it is relative to base_path. URL are
91 void path_combine(char *dest
, int dest_size
,
92 const char *base_path
,
100 if (path_is_absolute(filename
)) {
101 pstrcpy(dest
, dest_size
, filename
);
103 p
= strchr(base_path
, ':');
108 p1
= strrchr(base_path
, '/');
112 p2
= strrchr(base_path
, '\\');
124 if (len
> dest_size
- 1)
126 memcpy(dest
, base_path
, len
);
128 pstrcat(dest
, dest_size
, filename
);
132 void bdrv_register(BlockDriver
*bdrv
)
134 if (!bdrv
->bdrv_aio_readv
) {
135 /* add AIO emulation layer */
136 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
137 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
138 } else if (!bdrv
->bdrv_read
) {
139 /* add synchronous IO emulation layer */
140 bdrv
->bdrv_read
= bdrv_read_em
;
141 bdrv
->bdrv_write
= bdrv_write_em
;
144 if (!bdrv
->bdrv_aio_flush
)
145 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
147 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
150 /* create a new block device (by default it is empty) */
151 BlockDriverState
*bdrv_new(const char *device_name
)
153 BlockDriverState
*bs
;
155 bs
= qemu_mallocz(sizeof(BlockDriverState
));
156 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
157 if (device_name
[0] != '\0') {
158 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
163 BlockDriver
*bdrv_find_format(const char *format_name
)
166 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
167 if (!strcmp(drv1
->format_name
, format_name
)) {
174 static int bdrv_is_whitelisted(BlockDriver
*drv
)
176 static const char *whitelist
[] = {
177 CONFIG_BDRV_WHITELIST
182 return 1; /* no whitelist, anything goes */
184 for (p
= whitelist
; *p
; p
++) {
185 if (!strcmp(drv
->format_name
, *p
)) {
192 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
194 BlockDriver
*drv
= bdrv_find_format(format_name
);
195 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
198 int bdrv_create(BlockDriver
*drv
, const char* filename
,
199 QEMUOptionParameter
*options
)
201 if (!drv
->bdrv_create
)
204 return drv
->bdrv_create(filename
, options
);
207 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
211 drv
= find_protocol(filename
);
213 drv
= bdrv_find_format("file");
216 return bdrv_create(drv
, filename
, options
);
220 void get_tmp_filename(char *filename
, int size
)
222 char temp_dir
[MAX_PATH
];
224 GetTempPath(MAX_PATH
, temp_dir
);
225 GetTempFileName(temp_dir
, "qem", 0, filename
);
228 void get_tmp_filename(char *filename
, int size
)
232 /* XXX: race condition possible */
233 tmpdir
= getenv("TMPDIR");
236 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
237 fd
= mkstemp(filename
);
243 static int is_windows_drive_prefix(const char *filename
)
245 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
246 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
250 int is_windows_drive(const char *filename
)
252 if (is_windows_drive_prefix(filename
) &&
255 if (strstart(filename
, "\\\\.\\", NULL
) ||
256 strstart(filename
, "//./", NULL
))
263 * Detect host devices. By convention, /dev/cdrom[N] is always
264 * recognized as a host CDROM.
266 static BlockDriver
*find_hdev_driver(const char *filename
)
268 int score_max
= 0, score
;
269 BlockDriver
*drv
= NULL
, *d
;
271 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
272 if (d
->bdrv_probe_device
) {
273 score
= d
->bdrv_probe_device(filename
);
274 if (score
> score_max
) {
284 static BlockDriver
*find_protocol(const char *filename
)
292 /* TODO Drivers without bdrv_file_open must be specified explicitly */
295 is_drive
= is_windows_drive(filename
) ||
296 is_windows_drive_prefix(filename
);
300 p
= strchr(filename
, ':');
301 if (!p
|| is_drive
) {
302 drv1
= find_hdev_driver(filename
);
304 drv1
= bdrv_find_format("file");
309 if (len
> sizeof(protocol
) - 1)
310 len
= sizeof(protocol
) - 1;
311 memcpy(protocol
, filename
, len
);
312 protocol
[len
] = '\0';
313 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
314 if (drv1
->protocol_name
&&
315 !strcmp(drv1
->protocol_name
, protocol
)) {
322 static BlockDriver
*find_image_format(const char *filename
)
324 int ret
, score
, score_max
;
325 BlockDriver
*drv1
, *drv
;
327 BlockDriverState
*bs
;
329 ret
= bdrv_file_open(&bs
, filename
, 0);
333 /* Return the raw BlockDriver * to scsi-generic devices */
335 return bdrv_find_format("raw");
337 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
345 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
346 if (drv1
->bdrv_probe
) {
347 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
348 if (score
> score_max
) {
358 * Set the current 'total_sectors' value
360 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
362 BlockDriver
*drv
= bs
->drv
;
364 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
368 /* query actual device if possible, otherwise just trust the hint */
369 if (drv
->bdrv_getlength
) {
370 int64_t length
= drv
->bdrv_getlength(bs
);
374 hint
= length
>> BDRV_SECTOR_BITS
;
377 bs
->total_sectors
= hint
;
382 * Common part for opening disk images and files
384 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
385 int flags
, BlockDriver
*drv
)
392 bs
->total_sectors
= 0;
393 bs
->is_temporary
= 0;
396 bs
->open_flags
= flags
;
397 /* buffer_alignment defaulted to 512, drivers can change this value */
398 bs
->buffer_alignment
= 512;
400 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
402 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
407 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
410 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
411 * write cache to the guest. We do need the fdatasync to flush
412 * out transactions for block allocations, and we maybe have a
413 * volatile write cache in our backing device to deal with.
415 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
416 bs
->enable_write_cache
= 1;
419 * Clear flags that are internal to the block layer before opening the
422 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
425 * Snapshots should be writeable.
427 if (bs
->is_temporary
) {
428 open_flags
|= BDRV_O_RDWR
;
431 /* Open the image, either directly or using a protocol */
432 if (drv
->bdrv_file_open
) {
433 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
435 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
437 ret
= drv
->bdrv_open(bs
, open_flags
);
445 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
447 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
453 if (bs
->is_temporary
) {
461 bdrv_delete(bs
->file
);
464 qemu_free(bs
->opaque
);
471 * Opens a file using a protocol (file, host_device, nbd, ...)
473 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
475 BlockDriverState
*bs
;
479 drv
= find_protocol(filename
);
485 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
496 * Opens a disk image (raw, qcow2, vmdk, ...)
498 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
503 if (flags
& BDRV_O_SNAPSHOT
) {
504 BlockDriverState
*bs1
;
507 BlockDriver
*bdrv_qcow2
;
508 QEMUOptionParameter
*options
;
509 char tmp_filename
[PATH_MAX
];
510 char backing_filename
[PATH_MAX
];
512 /* if snapshot, we create a temporary backing file and open it
513 instead of opening 'filename' directly */
515 /* if there is a backing file, use it */
517 ret
= bdrv_open(bs1
, filename
, 0, drv
);
522 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
524 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
529 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
531 /* Real path is meaningless for protocols */
533 snprintf(backing_filename
, sizeof(backing_filename
),
535 else if (!realpath(filename
, backing_filename
))
538 bdrv_qcow2
= bdrv_find_format("qcow2");
539 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
541 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
542 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
544 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
548 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
549 free_option_parameters(options
);
554 filename
= tmp_filename
;
556 bs
->is_temporary
= 1;
559 /* Find the right image format driver */
561 drv
= find_image_format(filename
);
566 goto unlink_and_fail
;
570 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
572 goto unlink_and_fail
;
575 /* If there is a backing file, use it */
576 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
577 char backing_filename
[PATH_MAX
];
579 BlockDriver
*back_drv
= NULL
;
581 bs
->backing_hd
= bdrv_new("");
582 path_combine(backing_filename
, sizeof(backing_filename
),
583 filename
, bs
->backing_file
);
584 if (bs
->backing_format
[0] != '\0')
585 back_drv
= bdrv_find_format(bs
->backing_format
);
587 /* backing files always opened read-only */
589 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
591 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
596 if (bs
->is_temporary
) {
597 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
599 /* base image inherits from "parent" */
600 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
604 if (!bdrv_key_required(bs
)) {
605 /* call the change callback */
606 bs
->media_changed
= 1;
608 bs
->change_cb(bs
->change_opaque
);
614 if (bs
->is_temporary
) {
620 void bdrv_close(BlockDriverState
*bs
)
623 if (bs
->backing_hd
) {
624 bdrv_delete(bs
->backing_hd
);
625 bs
->backing_hd
= NULL
;
627 bs
->drv
->bdrv_close(bs
);
628 qemu_free(bs
->opaque
);
630 if (bs
->is_temporary
) {
631 unlink(bs
->filename
);
637 if (bs
->file
!= NULL
) {
638 bdrv_close(bs
->file
);
641 /* call the change callback */
642 bs
->media_changed
= 1;
644 bs
->change_cb(bs
->change_opaque
);
648 void bdrv_delete(BlockDriverState
*bs
)
650 /* remove from list, if necessary */
651 if (bs
->device_name
[0] != '\0') {
652 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
656 if (bs
->file
!= NULL
) {
657 bdrv_delete(bs
->file
);
664 * Run consistency checks on an image
666 * Returns the number of errors or -errno when an internal error occurs
668 int bdrv_check(BlockDriverState
*bs
)
670 if (bs
->drv
->bdrv_check
== NULL
) {
674 return bs
->drv
->bdrv_check(bs
);
677 /* commit COW file into the raw image */
678 int bdrv_commit(BlockDriverState
*bs
)
680 BlockDriver
*drv
= bs
->drv
;
681 int64_t i
, total_sectors
;
682 int n
, j
, ro
, open_flags
;
683 int ret
= 0, rw_ret
= 0;
684 unsigned char sector
[512];
686 BlockDriverState
*bs_rw
, *bs_ro
;
691 if (!bs
->backing_hd
) {
695 if (bs
->backing_hd
->keep_read_only
) {
699 ro
= bs
->backing_hd
->read_only
;
700 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
701 open_flags
= bs
->backing_hd
->open_flags
;
705 bdrv_delete(bs
->backing_hd
);
706 bs
->backing_hd
= NULL
;
707 bs_rw
= bdrv_new("");
708 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, drv
);
711 /* try to re-open read-only */
712 bs_ro
= bdrv_new("");
713 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, drv
);
716 /* drive not functional anymore */
720 bs
->backing_hd
= bs_ro
;
723 bs
->backing_hd
= bs_rw
;
726 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
727 for (i
= 0; i
< total_sectors
;) {
728 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
729 for(j
= 0; j
< n
; j
++) {
730 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
735 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
746 if (drv
->bdrv_make_empty
) {
747 ret
= drv
->bdrv_make_empty(bs
);
752 * Make sure all data we wrote to the backing device is actually
756 bdrv_flush(bs
->backing_hd
);
762 bdrv_delete(bs
->backing_hd
);
763 bs
->backing_hd
= NULL
;
764 bs_ro
= bdrv_new("");
765 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, drv
);
768 /* drive not functional anymore */
772 bs
->backing_hd
= bs_ro
;
773 bs
->backing_hd
->keep_read_only
= 0;
782 * -EINVAL - backing format specified, but no file
783 * -ENOSPC - can't update the backing file because no space is left in the
785 * -ENOTSUP - format driver doesn't support changing the backing file
787 int bdrv_change_backing_file(BlockDriverState
*bs
,
788 const char *backing_file
, const char *backing_fmt
)
790 BlockDriver
*drv
= bs
->drv
;
792 if (drv
->bdrv_change_backing_file
!= NULL
) {
793 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
799 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
804 if (!bdrv_is_inserted(bs
))
810 len
= bdrv_getlength(bs
);
815 if ((offset
> len
) || (len
- offset
< size
))
821 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
824 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
827 /* return < 0 if error. See bdrv_write() for the return codes */
828 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
829 uint8_t *buf
, int nb_sectors
)
831 BlockDriver
*drv
= bs
->drv
;
835 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
838 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
841 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
842 int nb_sectors
, int dirty
)
845 unsigned long val
, idx
, bit
;
847 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
848 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
850 for (; start
<= end
; start
++) {
851 idx
= start
/ (sizeof(unsigned long) * 8);
852 bit
= start
% (sizeof(unsigned long) * 8);
853 val
= bs
->dirty_bitmap
[idx
];
855 if (!(val
& (1 << bit
))) {
860 if (val
& (1 << bit
)) {
865 bs
->dirty_bitmap
[idx
] = val
;
869 /* Return < 0 if error. Important errors are:
870 -EIO generic I/O error (may happen for all errors)
871 -ENOMEDIUM No media inserted.
872 -EINVAL Invalid sector number or nb_sectors
873 -EACCES Trying to write a read-only device
875 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
876 const uint8_t *buf
, int nb_sectors
)
878 BlockDriver
*drv
= bs
->drv
;
883 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
886 if (bs
->dirty_bitmap
) {
887 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
890 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
891 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
894 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
897 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
898 void *buf
, int count1
)
900 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
901 int len
, nb_sectors
, count
;
906 /* first read to align to sector start */
907 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
910 sector_num
= offset
>> BDRV_SECTOR_BITS
;
912 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
914 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
922 /* read the sectors "in place" */
923 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
924 if (nb_sectors
> 0) {
925 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
927 sector_num
+= nb_sectors
;
928 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
933 /* add data from the last sector */
935 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
937 memcpy(buf
, tmp_buf
, count
);
942 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
943 const void *buf
, int count1
)
945 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
946 int len
, nb_sectors
, count
;
951 /* first write to align to sector start */
952 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
955 sector_num
= offset
>> BDRV_SECTOR_BITS
;
957 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
959 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
960 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
969 /* write the sectors "in place" */
970 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
971 if (nb_sectors
> 0) {
972 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
974 sector_num
+= nb_sectors
;
975 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
980 /* add data from the last sector */
982 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
984 memcpy(tmp_buf
, buf
, count
);
985 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
992 * Truncate file to 'offset' bytes (needed only for file protocols)
994 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
996 BlockDriver
*drv
= bs
->drv
;
1000 if (!drv
->bdrv_truncate
)
1004 ret
= drv
->bdrv_truncate(bs
, offset
);
1006 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1012 * Length of a file in bytes. Return < 0 if error or unknown.
1014 int64_t bdrv_getlength(BlockDriverState
*bs
)
1016 BlockDriver
*drv
= bs
->drv
;
1020 /* Fixed size devices use the total_sectors value for speed instead of
1021 issuing a length query (like lseek) on each call. Also, legacy block
1022 drivers don't provide a bdrv_getlength function and must use
1024 if (!bs
->growable
|| !drv
->bdrv_getlength
) {
1025 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1027 return drv
->bdrv_getlength(bs
);
1030 /* return 0 as number of sectors if no device present or error */
1031 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1034 length
= bdrv_getlength(bs
);
1038 length
= length
>> BDRV_SECTOR_BITS
;
1039 *nb_sectors_ptr
= length
;
1043 uint8_t boot_ind
; /* 0x80 - active */
1044 uint8_t head
; /* starting head */
1045 uint8_t sector
; /* starting sector */
1046 uint8_t cyl
; /* starting cylinder */
1047 uint8_t sys_ind
; /* What partition type */
1048 uint8_t end_head
; /* end head */
1049 uint8_t end_sector
; /* end sector */
1050 uint8_t end_cyl
; /* end cylinder */
1051 uint32_t start_sect
; /* starting sector counting from 0 */
1052 uint32_t nr_sects
; /* nr of sectors in partition */
1053 } __attribute__((packed
));
1055 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1056 static int guess_disk_lchs(BlockDriverState
*bs
,
1057 int *pcylinders
, int *pheads
, int *psectors
)
1060 int ret
, i
, heads
, sectors
, cylinders
;
1061 struct partition
*p
;
1063 uint64_t nb_sectors
;
1065 bdrv_get_geometry(bs
, &nb_sectors
);
1067 ret
= bdrv_read(bs
, 0, buf
, 1);
1070 /* test msdos magic */
1071 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1073 for(i
= 0; i
< 4; i
++) {
1074 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1075 nr_sects
= le32_to_cpu(p
->nr_sects
);
1076 if (nr_sects
&& p
->end_head
) {
1077 /* We make the assumption that the partition terminates on
1078 a cylinder boundary */
1079 heads
= p
->end_head
+ 1;
1080 sectors
= p
->end_sector
& 63;
1083 cylinders
= nb_sectors
/ (heads
* sectors
);
1084 if (cylinders
< 1 || cylinders
> 16383)
1087 *psectors
= sectors
;
1088 *pcylinders
= cylinders
;
1090 printf("guessed geometry: LCHS=%d %d %d\n",
1091 cylinders
, heads
, sectors
);
1099 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1101 int translation
, lba_detected
= 0;
1102 int cylinders
, heads
, secs
;
1103 uint64_t nb_sectors
;
1105 /* if a geometry hint is available, use it */
1106 bdrv_get_geometry(bs
, &nb_sectors
);
1107 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1108 translation
= bdrv_get_translation_hint(bs
);
1109 if (cylinders
!= 0) {
1114 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1116 /* if heads > 16, it means that a BIOS LBA
1117 translation was active, so the default
1118 hardware geometry is OK */
1120 goto default_geometry
;
1125 /* disable any translation to be in sync with
1126 the logical geometry */
1127 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1128 bdrv_set_translation_hint(bs
,
1129 BIOS_ATA_TRANSLATION_NONE
);
1134 /* if no geometry, use a standard physical disk geometry */
1135 cylinders
= nb_sectors
/ (16 * 63);
1137 if (cylinders
> 16383)
1139 else if (cylinders
< 2)
1144 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1145 if ((*pcyls
* *pheads
) <= 131072) {
1146 bdrv_set_translation_hint(bs
,
1147 BIOS_ATA_TRANSLATION_LARGE
);
1149 bdrv_set_translation_hint(bs
,
1150 BIOS_ATA_TRANSLATION_LBA
);
1154 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1158 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1159 int cyls
, int heads
, int secs
)
1166 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1169 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1170 type
== BDRV_TYPE_FLOPPY
));
1173 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1175 bs
->translation
= translation
;
1178 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1179 int *pcyls
, int *pheads
, int *psecs
)
1182 *pheads
= bs
->heads
;
1186 int bdrv_get_type_hint(BlockDriverState
*bs
)
1191 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1193 return bs
->translation
;
1196 int bdrv_is_removable(BlockDriverState
*bs
)
1198 return bs
->removable
;
1201 int bdrv_is_read_only(BlockDriverState
*bs
)
1203 return bs
->read_only
;
1206 int bdrv_is_sg(BlockDriverState
*bs
)
1211 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1213 return bs
->enable_write_cache
;
1216 /* XXX: no longer used */
1217 void bdrv_set_change_cb(BlockDriverState
*bs
,
1218 void (*change_cb
)(void *opaque
), void *opaque
)
1220 bs
->change_cb
= change_cb
;
1221 bs
->change_opaque
= opaque
;
1224 int bdrv_is_encrypted(BlockDriverState
*bs
)
1226 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1228 return bs
->encrypted
;
1231 int bdrv_key_required(BlockDriverState
*bs
)
1233 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1235 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1237 return (bs
->encrypted
&& !bs
->valid_key
);
1240 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1243 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1244 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1250 if (!bs
->encrypted
) {
1252 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1255 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1258 } else if (!bs
->valid_key
) {
1260 /* call the change callback now, we skipped it on open */
1261 bs
->media_changed
= 1;
1263 bs
->change_cb(bs
->change_opaque
);
1268 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1273 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1277 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1282 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1283 it(opaque
, drv
->format_name
);
1287 BlockDriverState
*bdrv_find(const char *name
)
1289 BlockDriverState
*bs
;
1291 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1292 if (!strcmp(name
, bs
->device_name
)) {
1299 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1301 BlockDriverState
*bs
;
1303 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1308 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1310 return bs
->device_name
;
1313 void bdrv_flush(BlockDriverState
*bs
)
1315 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1316 bs
->drv
->bdrv_flush(bs
);
1319 void bdrv_flush_all(void)
1321 BlockDriverState
*bs
;
1323 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1324 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1325 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1331 int bdrv_has_zero_init(BlockDriverState
*bs
)
1335 if (bs
->drv
->no_zero_init
) {
1337 } else if (bs
->file
) {
1338 return bdrv_has_zero_init(bs
->file
);
1345 * Returns true iff the specified sector is present in the disk image. Drivers
1346 * not implementing the functionality are assumed to not support backing files,
1347 * hence all their sectors are reported as allocated.
1349 * 'pnum' is set to the number of sectors (including and immediately following
1350 * the specified sector) that are known to be in the same
1351 * allocated/unallocated state.
1353 * 'nb_sectors' is the max value 'pnum' should be set to.
1355 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1359 if (!bs
->drv
->bdrv_is_allocated
) {
1360 if (sector_num
>= bs
->total_sectors
) {
1364 n
= bs
->total_sectors
- sector_num
;
1365 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1368 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1371 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1372 BlockMonEventAction action
, int is_read
)
1375 const char *action_str
;
1378 case BDRV_ACTION_REPORT
:
1379 action_str
= "report";
1381 case BDRV_ACTION_IGNORE
:
1382 action_str
= "ignore";
1384 case BDRV_ACTION_STOP
:
1385 action_str
= "stop";
1391 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1394 is_read
? "read" : "write");
1395 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1397 qobject_decref(data
);
1400 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1403 Monitor
*mon
= opaque
;
1405 bs_dict
= qobject_to_qdict(obj
);
1407 monitor_printf(mon
, "%s: type=%s removable=%d",
1408 qdict_get_str(bs_dict
, "device"),
1409 qdict_get_str(bs_dict
, "type"),
1410 qdict_get_bool(bs_dict
, "removable"));
1412 if (qdict_get_bool(bs_dict
, "removable")) {
1413 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1416 if (qdict_haskey(bs_dict
, "inserted")) {
1417 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1419 monitor_printf(mon
, " file=");
1420 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1421 if (qdict_haskey(qdict
, "backing_file")) {
1422 monitor_printf(mon
, " backing_file=");
1423 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1425 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1426 qdict_get_bool(qdict
, "ro"),
1427 qdict_get_str(qdict
, "drv"),
1428 qdict_get_bool(qdict
, "encrypted"));
1430 monitor_printf(mon
, " [not inserted]");
1433 monitor_printf(mon
, "\n");
1436 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1438 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1442 * bdrv_info(): Block devices information
1444 * Each block device information is stored in a QDict and the
1445 * returned QObject is a QList of all devices.
1447 * The QDict contains the following:
1449 * - "device": device name
1450 * - "type": device type
1451 * - "removable": true if the device is removable, false otherwise
1452 * - "locked": true if the device is locked, false otherwise
1453 * - "inserted": only present if the device is inserted, it is a QDict
1454 * containing the following:
1455 * - "file": device file name
1456 * - "ro": true if read-only, false otherwise
1457 * - "drv": driver format name
1458 * - "backing_file": backing file name if one is used
1459 * - "encrypted": true if encrypted, false otherwise
1463 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1464 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1465 * { "device": "floppy0", "type": "floppy", "removable": true,
1466 * "locked": false } ]
1468 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1471 BlockDriverState
*bs
;
1473 bs_list
= qlist_new();
1475 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1477 const char *type
= "unknown";
1483 case BDRV_TYPE_CDROM
:
1486 case BDRV_TYPE_FLOPPY
:
1491 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1492 "'removable': %i, 'locked': %i }",
1493 bs
->device_name
, type
, bs
->removable
,
1498 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1500 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1501 "'encrypted': %i }",
1502 bs
->filename
, bs
->read_only
,
1503 bs
->drv
->format_name
,
1504 bdrv_is_encrypted(bs
));
1505 if (bs
->backing_file
[0] != '\0') {
1506 QDict
*qdict
= qobject_to_qdict(obj
);
1507 qdict_put(qdict
, "backing_file",
1508 qstring_from_str(bs
->backing_file
));
1511 qdict_put_obj(bs_dict
, "inserted", obj
);
1513 qlist_append_obj(bs_list
, bs_obj
);
1516 *ret_data
= QOBJECT(bs_list
);
1519 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1522 Monitor
*mon
= opaque
;
1524 qdict
= qobject_to_qdict(data
);
1525 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1527 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1528 monitor_printf(mon
, " rd_bytes=%" PRId64
1529 " wr_bytes=%" PRId64
1530 " rd_operations=%" PRId64
1531 " wr_operations=%" PRId64
1533 qdict_get_int(qdict
, "rd_bytes"),
1534 qdict_get_int(qdict
, "wr_bytes"),
1535 qdict_get_int(qdict
, "rd_operations"),
1536 qdict_get_int(qdict
, "wr_operations"));
1539 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1541 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1544 static QObject
* bdrv_info_stats_bs(BlockDriverState
*bs
)
1549 res
= qobject_from_jsonf("{ 'stats': {"
1550 "'rd_bytes': %" PRId64
","
1551 "'wr_bytes': %" PRId64
","
1552 "'rd_operations': %" PRId64
","
1553 "'wr_operations': %" PRId64
","
1554 "'wr_highest_offset': %" PRId64
1556 bs
->rd_bytes
, bs
->wr_bytes
,
1557 bs
->rd_ops
, bs
->wr_ops
,
1558 bs
->wr_highest_sector
* 512);
1559 dict
= qobject_to_qdict(res
);
1561 if (*bs
->device_name
) {
1562 qdict_put(dict
, "device", qstring_from_str(bs
->device_name
));
1566 QObject
*parent
= bdrv_info_stats_bs(bs
->file
);
1567 qdict_put_obj(dict
, "parent", parent
);
1574 * bdrv_info_stats(): show block device statistics
1576 * Each device statistic information is stored in a QDict and
1577 * the returned QObject is a QList of all devices.
1579 * The QDict contains the following:
1581 * - "device": device name
1582 * - "stats": A QDict with the statistics information, it contains:
1583 * - "rd_bytes": bytes read
1584 * - "wr_bytes": bytes written
1585 * - "rd_operations": read operations
1586 * - "wr_operations": write operations
1587 * - "wr_highest_offset": Highest offset of a sector written since the
1588 * BlockDriverState has been opened
1589 * - "parent": A QDict recursively holding the statistics of the underlying
1590 * protocol (e.g. the host file for a qcow2 image). If there is no
1591 * underlying protocol, this field is omitted.
1595 * [ { "device": "ide0-hd0",
1596 * "stats": { "rd_bytes": 512,
1598 * "rd_operations": 1,
1599 * "wr_operations": 0,
1600 * "wr_highest_offset": 0 },
1602 * "stats": { "rd_bytes": 1024,
1604 * "rd_operations": 2,
1605 * "wr_operations": 0,
1606 * "wr_highest_offset": 0,
1608 * { "device": "ide1-cd0",
1609 * "stats": { "rd_bytes": 0,
1611 * "rd_operations": 0,
1612 * "wr_operations": 0,
1613 * "wr_highest_offset": 0 } },
1615 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1619 BlockDriverState
*bs
;
1621 devices
= qlist_new();
1623 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1624 obj
= bdrv_info_stats_bs(bs
);
1625 qlist_append_obj(devices
, obj
);
1628 *ret_data
= QOBJECT(devices
);
1631 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1633 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1634 return bs
->backing_file
;
1635 else if (bs
->encrypted
)
1636 return bs
->filename
;
1641 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1642 char *filename
, int filename_size
)
1644 if (!bs
->backing_file
) {
1645 pstrcpy(filename
, filename_size
, "");
1647 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1651 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1652 const uint8_t *buf
, int nb_sectors
)
1654 BlockDriver
*drv
= bs
->drv
;
1657 if (!drv
->bdrv_write_compressed
)
1659 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1662 if (bs
->dirty_bitmap
) {
1663 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1666 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1669 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1671 BlockDriver
*drv
= bs
->drv
;
1674 if (!drv
->bdrv_get_info
)
1676 memset(bdi
, 0, sizeof(*bdi
));
1677 return drv
->bdrv_get_info(bs
, bdi
);
1680 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1681 int64_t pos
, int size
)
1683 BlockDriver
*drv
= bs
->drv
;
1686 if (!drv
->bdrv_save_vmstate
)
1688 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1691 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1692 int64_t pos
, int size
)
1694 BlockDriver
*drv
= bs
->drv
;
1697 if (!drv
->bdrv_load_vmstate
)
1699 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1702 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1704 BlockDriver
*drv
= bs
->drv
;
1706 if (!drv
|| !drv
->bdrv_debug_event
) {
1710 return drv
->bdrv_debug_event(bs
, event
);
1714 /**************************************************************/
1715 /* handling of snapshots */
1717 int bdrv_snapshot_create(BlockDriverState
*bs
,
1718 QEMUSnapshotInfo
*sn_info
)
1720 BlockDriver
*drv
= bs
->drv
;
1723 if (!drv
->bdrv_snapshot_create
)
1725 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1728 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1729 const char *snapshot_id
)
1731 BlockDriver
*drv
= bs
->drv
;
1734 if (!drv
->bdrv_snapshot_goto
)
1736 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1739 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1741 BlockDriver
*drv
= bs
->drv
;
1744 if (!drv
->bdrv_snapshot_delete
)
1746 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1749 int bdrv_snapshot_list(BlockDriverState
*bs
,
1750 QEMUSnapshotInfo
**psn_info
)
1752 BlockDriver
*drv
= bs
->drv
;
1755 if (!drv
->bdrv_snapshot_list
)
1757 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1760 #define NB_SUFFIXES 4
1762 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1764 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1769 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1772 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1773 if (size
< (10 * base
)) {
1774 snprintf(buf
, buf_size
, "%0.1f%c",
1775 (double)size
/ base
,
1778 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1779 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1780 ((size
+ (base
>> 1)) / base
),
1790 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1792 char buf1
[128], date_buf
[128], clock_buf
[128];
1802 snprintf(buf
, buf_size
,
1803 "%-10s%-20s%7s%20s%15s",
1804 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1808 ptm
= localtime(&ti
);
1809 strftime(date_buf
, sizeof(date_buf
),
1810 "%Y-%m-%d %H:%M:%S", ptm
);
1812 localtime_r(&ti
, &tm
);
1813 strftime(date_buf
, sizeof(date_buf
),
1814 "%Y-%m-%d %H:%M:%S", &tm
);
1816 secs
= sn
->vm_clock_nsec
/ 1000000000;
1817 snprintf(clock_buf
, sizeof(clock_buf
),
1818 "%02d:%02d:%02d.%03d",
1820 (int)((secs
/ 60) % 60),
1822 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1823 snprintf(buf
, buf_size
,
1824 "%-10s%-20s%7s%20s%15s",
1825 sn
->id_str
, sn
->name
,
1826 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1834 /**************************************************************/
1837 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1838 QEMUIOVector
*qiov
, int nb_sectors
,
1839 BlockDriverCompletionFunc
*cb
, void *opaque
)
1841 BlockDriver
*drv
= bs
->drv
;
1842 BlockDriverAIOCB
*ret
;
1846 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1849 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1853 /* Update stats even though technically transfer has not happened. */
1854 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1861 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1862 QEMUIOVector
*qiov
, int nb_sectors
,
1863 BlockDriverCompletionFunc
*cb
, void *opaque
)
1865 BlockDriver
*drv
= bs
->drv
;
1866 BlockDriverAIOCB
*ret
;
1872 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1875 if (bs
->dirty_bitmap
) {
1876 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1879 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1883 /* Update stats even though technically transfer has not happened. */
1884 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1886 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
1887 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
1895 typedef struct MultiwriteCB
{
1900 BlockDriverCompletionFunc
*cb
;
1902 QEMUIOVector
*free_qiov
;
1907 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1911 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1912 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1913 if (mcb
->callbacks
[i
].free_qiov
) {
1914 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
1916 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1917 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1921 static void multiwrite_cb(void *opaque
, int ret
)
1923 MultiwriteCB
*mcb
= opaque
;
1925 if (ret
< 0 && !mcb
->error
) {
1927 multiwrite_user_cb(mcb
);
1930 mcb
->num_requests
--;
1931 if (mcb
->num_requests
== 0) {
1932 if (mcb
->error
== 0) {
1933 multiwrite_user_cb(mcb
);
1939 static int multiwrite_req_compare(const void *a
, const void *b
)
1941 const BlockRequest
*req1
= a
, *req2
= b
;
1944 * Note that we can't simply subtract req2->sector from req1->sector
1945 * here as that could overflow the return value.
1947 if (req1
->sector
> req2
->sector
) {
1949 } else if (req1
->sector
< req2
->sector
) {
1957 * Takes a bunch of requests and tries to merge them. Returns the number of
1958 * requests that remain after merging.
1960 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1961 int num_reqs
, MultiwriteCB
*mcb
)
1965 // Sort requests by start sector
1966 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1968 // Check if adjacent requests touch the same clusters. If so, combine them,
1969 // filling up gaps with zero sectors.
1971 for (i
= 1; i
< num_reqs
; i
++) {
1973 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1975 // This handles the cases that are valid for all block drivers, namely
1976 // exactly sequential writes and overlapping writes.
1977 if (reqs
[i
].sector
<= oldreq_last
) {
1981 // The block driver may decide that it makes sense to combine requests
1982 // even if there is a gap of some sectors between them. In this case,
1983 // the gap is filled with zeros (therefore only applicable for yet
1984 // unused space in format like qcow2).
1985 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1986 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1989 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
1995 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1996 qemu_iovec_init(qiov
,
1997 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1999 // Add the first request to the merged one. If the requests are
2000 // overlapping, drop the last sectors of the first request.
2001 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
2002 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
2004 // We might need to add some zeros between the two requests
2005 if (reqs
[i
].sector
> oldreq_last
) {
2006 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
2007 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
2008 memset(buf
, 0, zero_bytes
);
2009 qemu_iovec_add(qiov
, buf
, zero_bytes
);
2010 mcb
->callbacks
[i
].free_buf
= buf
;
2013 // Add the second request
2014 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2016 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
2017 reqs
[outidx
].qiov
= qiov
;
2019 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2022 reqs
[outidx
].sector
= reqs
[i
].sector
;
2023 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2024 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2032 * Submit multiple AIO write requests at once.
2034 * On success, the function returns 0 and all requests in the reqs array have
2035 * been submitted. In error case this function returns -1, and any of the
2036 * requests may or may not be submitted yet. In particular, this means that the
2037 * callback will be called for some of the requests, for others it won't. The
2038 * caller must check the error field of the BlockRequest to wait for the right
2039 * callbacks (if error != 0, no callback will be called).
2041 * The implementation may modify the contents of the reqs array, e.g. to merge
2042 * requests. However, the fields opaque and error are left unmodified as they
2043 * are used to signal failure for a single request to the caller.
2045 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2047 BlockDriverAIOCB
*acb
;
2051 if (num_reqs
== 0) {
2055 // Create MultiwriteCB structure
2056 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2057 mcb
->num_requests
= 0;
2058 mcb
->num_callbacks
= num_reqs
;
2060 for (i
= 0; i
< num_reqs
; i
++) {
2061 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2062 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2065 // Check for mergable requests
2066 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2068 // Run the aio requests
2069 for (i
= 0; i
< num_reqs
; i
++) {
2070 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2071 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2074 // We can only fail the whole thing if no request has been
2075 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2076 // complete and report the error in the callback.
2077 if (mcb
->num_requests
== 0) {
2078 reqs
[i
].error
= -EIO
;
2081 mcb
->num_requests
++;
2082 multiwrite_cb(mcb
, -EIO
);
2086 mcb
->num_requests
++;
2097 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2098 BlockDriverCompletionFunc
*cb
, void *opaque
)
2100 BlockDriver
*drv
= bs
->drv
;
2104 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2107 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2109 acb
->pool
->cancel(acb
);
2113 /**************************************************************/
2114 /* async block device emulation */
2116 typedef struct BlockDriverAIOCBSync
{
2117 BlockDriverAIOCB common
;
2120 /* vector translation state */
2124 } BlockDriverAIOCBSync
;
2126 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2128 BlockDriverAIOCBSync
*acb
=
2129 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2130 qemu_bh_delete(acb
->bh
);
2132 qemu_aio_release(acb
);
2135 static AIOPool bdrv_em_aio_pool
= {
2136 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2137 .cancel
= bdrv_aio_cancel_em
,
2140 static void bdrv_aio_bh_cb(void *opaque
)
2142 BlockDriverAIOCBSync
*acb
= opaque
;
2145 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2146 qemu_vfree(acb
->bounce
);
2147 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2148 qemu_bh_delete(acb
->bh
);
2150 qemu_aio_release(acb
);
2153 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2157 BlockDriverCompletionFunc
*cb
,
2162 BlockDriverAIOCBSync
*acb
;
2164 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2165 acb
->is_write
= is_write
;
2167 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2170 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2173 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2174 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2176 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2179 qemu_bh_schedule(acb
->bh
);
2181 return &acb
->common
;
2184 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2185 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2186 BlockDriverCompletionFunc
*cb
, void *opaque
)
2188 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2191 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2192 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2193 BlockDriverCompletionFunc
*cb
, void *opaque
)
2195 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2198 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2199 BlockDriverCompletionFunc
*cb
, void *opaque
)
2201 BlockDriverAIOCBSync
*acb
;
2203 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2204 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2210 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2213 qemu_bh_schedule(acb
->bh
);
2214 return &acb
->common
;
2217 /**************************************************************/
2218 /* sync block device emulation */
2220 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2222 *(int *)opaque
= ret
;
2225 #define NOT_DONE 0x7fffffff
2227 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2228 uint8_t *buf
, int nb_sectors
)
2231 BlockDriverAIOCB
*acb
;
2235 async_context_push();
2237 async_ret
= NOT_DONE
;
2238 iov
.iov_base
= (void *)buf
;
2239 iov
.iov_len
= nb_sectors
* 512;
2240 qemu_iovec_init_external(&qiov
, &iov
, 1);
2241 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2242 bdrv_rw_em_cb
, &async_ret
);
2248 while (async_ret
== NOT_DONE
) {
2254 async_context_pop();
2258 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2259 const uint8_t *buf
, int nb_sectors
)
2262 BlockDriverAIOCB
*acb
;
2266 async_context_push();
2268 async_ret
= NOT_DONE
;
2269 iov
.iov_base
= (void *)buf
;
2270 iov
.iov_len
= nb_sectors
* 512;
2271 qemu_iovec_init_external(&qiov
, &iov
, 1);
2272 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2273 bdrv_rw_em_cb
, &async_ret
);
2278 while (async_ret
== NOT_DONE
) {
2283 async_context_pop();
2287 void bdrv_init(void)
2289 module_call_init(MODULE_INIT_BLOCK
);
2292 void bdrv_init_with_whitelist(void)
2294 use_bdrv_whitelist
= 1;
2298 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2299 BlockDriverCompletionFunc
*cb
, void *opaque
)
2301 BlockDriverAIOCB
*acb
;
2303 if (pool
->free_aiocb
) {
2304 acb
= pool
->free_aiocb
;
2305 pool
->free_aiocb
= acb
->next
;
2307 acb
= qemu_mallocz(pool
->aiocb_size
);
2312 acb
->opaque
= opaque
;
2316 void qemu_aio_release(void *p
)
2318 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2319 AIOPool
*pool
= acb
->pool
;
2320 acb
->next
= pool
->free_aiocb
;
2321 pool
->free_aiocb
= acb
;
2324 /**************************************************************/
2325 /* removable device support */
2328 * Return TRUE if the media is present
2330 int bdrv_is_inserted(BlockDriverState
*bs
)
2332 BlockDriver
*drv
= bs
->drv
;
2336 if (!drv
->bdrv_is_inserted
)
2338 ret
= drv
->bdrv_is_inserted(bs
);
2343 * Return TRUE if the media changed since the last call to this
2344 * function. It is currently only used for floppy disks
2346 int bdrv_media_changed(BlockDriverState
*bs
)
2348 BlockDriver
*drv
= bs
->drv
;
2351 if (!drv
|| !drv
->bdrv_media_changed
)
2354 ret
= drv
->bdrv_media_changed(bs
);
2355 if (ret
== -ENOTSUP
)
2356 ret
= bs
->media_changed
;
2357 bs
->media_changed
= 0;
2362 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2364 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2366 BlockDriver
*drv
= bs
->drv
;
2373 if (!drv
|| !drv
->bdrv_eject
) {
2376 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2378 if (ret
== -ENOTSUP
) {
2387 int bdrv_is_locked(BlockDriverState
*bs
)
2393 * Lock or unlock the media (if it is locked, the user won't be able
2394 * to eject it manually).
2396 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2398 BlockDriver
*drv
= bs
->drv
;
2400 bs
->locked
= locked
;
2401 if (drv
&& drv
->bdrv_set_locked
) {
2402 drv
->bdrv_set_locked(bs
, locked
);
2406 /* needed for generic scsi interface */
2408 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2410 BlockDriver
*drv
= bs
->drv
;
2412 if (drv
&& drv
->bdrv_ioctl
)
2413 return drv
->bdrv_ioctl(bs
, req
, buf
);
2417 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2418 unsigned long int req
, void *buf
,
2419 BlockDriverCompletionFunc
*cb
, void *opaque
)
2421 BlockDriver
*drv
= bs
->drv
;
2423 if (drv
&& drv
->bdrv_aio_ioctl
)
2424 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2430 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2432 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2435 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2437 int64_t bitmap_size
;
2439 bs
->dirty_count
= 0;
2441 if (!bs
->dirty_bitmap
) {
2442 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2443 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2444 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2446 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2449 if (bs
->dirty_bitmap
) {
2450 qemu_free(bs
->dirty_bitmap
);
2451 bs
->dirty_bitmap
= NULL
;
2456 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2458 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2460 if (bs
->dirty_bitmap
&&
2461 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2462 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2463 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2469 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2472 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2475 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2477 return bs
->dirty_count
;